Samsung has developed deepfake technology that is able to take a single image and bring it to life, according to a paper published by Samsung’s Moscow-based researchers.
Using machine learning and artificial intelligence (AI) the researchers have taught a connectionist system to study facial expressions and then utilizes the learned information to animate paintings, photos or videos.
The computer applies a number of different facial expressions over the source image, transforming the photo into a life-like person. The technology only requires a single image, but the more you provide, the more realistic the result is as the system will have more poses, facial expressions, and angles to work with.
Images of renowned people such as Albert Einstein and Marilyn Monroe have been used to demonstrate the animation capabilities of this AI technology. Several potential models of the Mona Lisa have also been created, but developers say a realist animation is far more easily achieved using photographs, rather than paintings.
The technology uses a Generative Adversarial Network (or GAN), which works by splitting the AI‘s workload into separate parts. One set of algorithms, a generator, tries to create something which in this case is a human face while another set, a judge (or discriminator), tries to determine if it’s a real image or a fake one. If the judge determines the image is fake, the generator starts over again. If the judge is fooled into thinking the image is real, then the results are checked and evaluated if the algorithms need changing.
US technology company, NVIDIA, has also been using Generative Adversarial Networks to produce ultra-real faces generated by artificial intelligence.
The AI developments of creating ultra-real images have raised concerns on whether the technology could be utilized for malicious purposes with some saying the technology could be used for fake news and propaganda.